An Improved Transformer-Based Neural Machine Translation Strategy: Interacting-Head Attention

Table 9

Overall evaluation scores of interacting-head attention on IWSLT16 DE-EN evaluation set and test set.

Model	Dataset	Subset	Number of heads/head size
			32/16	64/8
Interacting-head attention	IWSLT16	dev	24.38	19.98
		Test	22.85	18.54

Note. The unit of the performance is BLEU.